Bringing AI to Life: Practical Steps for Associations

In this episode of Associations NOW Presents, guest host Joanna Pineda of the Associations Thrive podcast chats with Rob Wenger, CEO of Higher Logic, and Amanda DeLuke, senior privacy analyst at Higher Logic. The conversation unpacks Higher Logic's own AI adoption journey, starting in October 2023, and the strategies they used to engage employees with AI tools. Rob and Amanda share practical insights on how associations can begin their AI journey, create policies, and leverage AI tools while prioritizing data privacy. Whether you're just getting started or looking to expand your AI capabilities, this episode offers actionable advice on using AI to enhance operations, improve member engagement, and drive innovation.

Check out the video podcast here:

https://www.youtube.com/watch?v=JLKUv2GFpUY
 

This episode is sponsored by Higher Logic.

Associations NOW Presents is produced by Association Briefings.

 

Transcript

Joanna Pineda: [00:00:00] Hey everyone, I'm Joanna Pineda, CEO and Chief Troublemaker at Matrix Group International and host of the Associations Thrive podcast. Each week I interview and association our non profit CEO about their personal journey and the things their organization is doing to thrive. But today, I am thrilled to be a guest host for the ASAE podcast for Associations Now.

And I am absolutely honored because I'm going to be interviewing Rob Wenger. CEO of HigherLogic and Amanda DeLuke, Senior Privacy Analyst. We'd like to thank our episode sponsor, HigherLogic, for their support of this podcast. Rob and Amanda, welcome to the show.

Rob Wenger: Thank you. It's great to be here.

Joanna Pineda: Hey Rob, there's probably two people out there in the universe that don't know HigherLogic, and for those two people, tell us about HigherLogic.

Rob Wenger: So our main product is online community. It's a place where associations members can talk to each other, share documents, ask questions, network, all the kinds of good stuff that they might do in person. They do it online 365 days a year. And we also make marketing automation products and a bunch of other different add-ons to it, like volunteer management.

Mentor management, a lot of different things that really are geared toward helping associations and their members connect.

Joanna Pineda: Amanda, you've got an interesting title. Your title is Senior Privacy Analyst. What is a Senior Privacy Analyst at HgherLogic?

Amanda DeLuke: Yeah, so I am definitely someone that works across the company.

You know, privacy is really everyone's problem, right? Everyone works together, creating privacy champions. So I actually report to the legal team. I deal with audit and making sure we are compliant for any applicable privacy laws, also working with our sales team and our data processing agreement and working with marketing.

So really just collaboration across the company, just to make sure we're staying safe. And we're adhering to any applicable [00:02:00] laws and staying compliant. Well, that's

Joanna Pineda: really interesting because today our topic is AI. And Robyn and Vanda, I'm really curious to learn almost two sides of the AI journey at HigherLogic because I bet you're integrating AI into HigherLogic products, but I'm also really curious to learn about your AI journey as a company.

Like how do you get people trained up and excited about AI so that now it becomes part of the culture? So what do you want to start with?

Rob Wenger: Yeah, so we sort of kicked off our internal AI journey in October 2023. I had this idea and I guess in August, Hey, let's get everyone in the company using AI like for a day, like sort of one of those hackathon type thing.

And then report back on what they did that was cool. And so I started kicking this thing off and then the folks in legal and Amanda's team said, Hey, hey, let us. Think about this privacy stuff first. Don't just jump into this. So we spent a little bit of [00:03:00] time putting it together. We ended up doing a whole month.

AI October we called it.

Joanna Pineda: And this was in 2023.

Rob Wenger: 2023, right, so October 2023. And so we spent a bunch of time, Amanda's team spent a bunch of time coming up with sort of the rules of the thing. Like don't just go use any tool and don't put customer data in it and all these kinds of great things. She developed processes and stuff like that.

I'll let her talk about that because she knows way more about it than I do, but then we ended up doing it as a full month, and the idea was everyone spend 16 hours in October. Do some stuff, get in the teams of one, two or three, play with these tools and come back and report back. And what we ended up doing is having on Halloween, it was a contest with costumes and with who had the best usage of AI.

And so it was a lot of fun, but it was a big journey to get there. And I'll let Amanda talk about that because she did most of the work. I just have the big idea and she made it happen.

Amanda DeLuke: Yeah, so it's really important to establish an internal policy and, you [00:04:00] know, it's really kind of putting the guardrails on certain areas.

And what I see in legislation is they really focus on balancing innovation and safety. And I think that's what we're kind of doing here too, right? We want to innovate, but we want to do it safely. So putting in place those policies is really important and again, like collaborating with teams on what type of data do we have and what are we holding and what are we okay with as it relates to AI and what type of data goes into AI.

And I think Rob, you bring up such a good point because we talk a lot about testing and really getting our hands dirty. And I think that's so important and it's really important because AI is ever evolving, that we're always evolving our policy and evolving how we work internally as well.

Joanna Pineda: Well, let's step back here because I'm curious about this and I'll give you a little bit of background.

So I was at a luncheon with a bunch of association CEOs and a number of them said, [00:05:00] gosh, we're struggling with AI. What does that mean? And they said, we've asked our staff to use the tools and they're either getting pushback or they're not seeing the fruits of some of this experimentation. And so what would you say to an exec who says, look, I think AI can help my team be more productive, but how do they kick it off?

Like, how do you get people comfortable with the tools so that they are seeing the fruits of this amazing technology?

Rob Wenger: Couple of thoughts on that. I get some really good advice. Yes. About a year ago from an AI luminary and the gist of it was If you're trying to turn people's daily routine around, that's difficult.

It takes time, and it's difficult to get them to start. Once they start, it takes time to get it to work, right? And the advice was start higher up. Start at system level stuff, like get AI. Doing things that don't involve humans yet. I mean, humans, you want them in the loop, but it's sort of like automating things [00:06:00] as opposed to like using ChatGPT every day on your desktop, right?

So we really started there looking at all right, we're going to build a data warehouse. It's going to have all this data from around the company in it. And Amanda's part of that as to how we handle the stuff that really is like privacy related or would be data for clients. We don't really put that in there.

We put data about clients more, but keeping the privacy stuff working well, but then doing it on top of that to give us sort of answers as opposed to letting a person go into ChatGPT and just. Ask a whole bunch of questions. So starting at that end of the spectrum, you can get a lot more of those kinds of things done.

And, and, and I know we're gonna talk about the product later, but that's how we're thinking about the product as well. It's more about automation than it is about what you're used to seeing in JATTPT, for example. We have that functionality. But where we think the real bang for the buck is going to come from is having AIs do a lot of automated type tasks.

So we started there. And [00:07:00] then, of course, we do encourage our staff to use these tools. The easiest place for us to start was with our engineers. Some of the reason is because it's way behind the scenes. Like, they don't really touch client data. And they also, products of an AI are Then test it, right?

Because that's the whole point of software development. You have to write the software and develop it and test it. And if you have any, I do that work, you know, it's going to be right because it's got to go through that process. And so that is an easy place to start. They're also very technical. It's easy to get them going where it starts to become, you know, more interesting.

I'm not gonna say challenging because the uptakes been really good. But when you get into finance or you get into, you know, accounting and some of these other aspects of the business where it's not obvious, How the data is going to be used. You got to be careful, but you can see a lot of gains. So it's been quite a journey.

Like I say, we started almost a year and a half ago and it's become something where it is throughout the company. One of the thing we did in that timeframe is we actually appointed, [00:08:00] I call him the AI Czar, his actual title is chief data officer, but Steve was the head of engineering and it seems like we have a bunch of engineers, so it was like, we really need to get AI done right in the company. And so Steve has taken on the role of really shepherding it through and we brought in some now very specific AI tools. One we really like called Glean, which brings in data from across the company and then you can query it just like ChatGPT, but it's all internal to the company.

So that's really cool. And so he's been like And been pushing those kinds of things and working with Amanda's team to keep us compliant and to keep us safe. And her recommendations are also really super helpful for approaches that we take on these things. Well,

Joanna Pineda: let me ask you a question because most associations don't have the ability to just spin up a data warehouse.

You know, and say, wow, let's automate stuff. So for an association exec or a CEO who says, you know, I want to get people using these tools. How would you recommend that they put on an AI October or [00:09:00] an AI May to get people comfortable? So let's start with the policy. And Amanda, I agree with you that we start with policy, but I have clients that I've been.

thinking about their policy for a year, and in the meantime, we can't even record Zoom meetings. And nobody can use any AI tools. So I'm not sure that that's the solution. So how do you get a policy up and running fast so that you can get some stuff done, but do it safely?

Amanda DeLuke: Right. Yeah, I think again, it's going back to the data classification, right?

So you're gonna kind of use that as it relates to any laws that are applicable to you, or maybe the sensitivity of the data. So if you know that you have public data, right, or something, maybe even some internal data that is okay to be used within AI, start there. We know it's public data. We're using a public model.

We're okay with it. We let's do this in a low risk activity. Let's test it. I think there's a lot of ways to really get your hands on it without having to move up through and start using it with a high risk type of data. So I think that's the great approach here when you have a [00:10:00] AI policy to kind of leverage that type of data and where are you using it and.

It really does kind of link into the EU AI Act and other legislation that is actually banning and have like prohibited uses for any high risk data that you might put in there. So I think it's really good to focus on the general use of AI. So I think that's a way that you can really leverage getting that policy out and being able to use it very quickly.

Rob Wenger: You're going to use tools, right? I don't expect any association to create their own. LLMs or anything like that, right? You're going to use something else. And so, Amanda, I know you and your team evaluated every product that was being used in our AIA October thing and still have the approval process for any new product we use.

And so, as an example, you know, Copilot, which is Microsoft's product, and you can get that with the Office Suite. I think it's an add on, you probably know more than I do, but, but like that one, we've got a guarantee for Microsoft and none of the data [00:11:00] we put in there. And so that's an easy place to start, right?

Joanna Pineda: And Google says the same thing about Gemini.

Amanda DeLuke: Exactly, Rob. You definitely want to check the terms to make sure, if it's a private model and, you know, your data is not going to be trained on that, then it's not going to go out to the third party to be used, that data to be used. Then you're okay. Then you have those set guardrails on there and you go in with eyes wide open.

Like, okay, they're saying that this is a private model. And if it is being trained, that information is only private to you and not, being sent externally, right, or being used externally.

Rob Wenger: So we, we, we sort of think of that as a process or a procedure or guardrails is a great term for it, where we have that outlined and they did that 18 months ago, whatever it was, and we've been following it.

And then. You know, I don't know how often, daily, weekly, new things come up and the team is evaluating them and saying yes or no, right? Sometimes it takes a little time because it's not so obviously, you know, I always want to use the latest and greatest tools and the latest and greatest tools don't always have [00:12:00] the best.

Documentation on how they're getting it, so, uh, I mean, I just gotta spend a little more time to evaluate things like that, but I think that's the first step, I mean, you know, associations, they could be doing that today, even if they're not quite ready to use or jump in with both feet, putting these guardrails, because people are going to use it, right?

People are using ChatGPT and Gemni and, and all these sorts of tools, and so you want to make sure that they're using them safely.

Joanna Pineda: So what I'm hearing is, decide that you're going to have AI as a policy initiative, right, or a program initiative. Develop some policies, start simple, have some approved tools, and then give people explicit direction about using AI for specific use cases.

And give them, it sounds like you even gave them guidance about how much time to spend, so that then they either weren't spending too little time or too much time. on their AI exploration. So it sounds like what I'm hearing from you is be very explicit and intentional with the AI journey. And then at the end of it, you [00:13:00] had sharing.

So can you share some cool stuff that came out of AI October, if it's not too confidential?

Rob Wenger: There was nothing confidential. There were some amazing costumes as well, but, but yeah, it was actually the one who won the thing. I mean, she's super smart. So I'm not, I'm not saying I wouldn't have predicted it, but it was in finance.

Like she took data. And use some tools to make some really cool spreadsheets and reports out of it that took her manually every month. It took her, I want to say, two days to come up with these things. And then using this new process that mixed AI and some, you know, kind of like workflow tech tools. She got that down to a few hours.

And so, Wow. She won because, first of all, she's a finance person, but like, The technology she was able to employ was very impressive and, you know, literally saving two days a month. That's life changing. Yes, it is. And for her, especially, because she's like, well, now I don't have to do those two days work, right?

[00:14:00] She can do other things.

Joanna Pineda: She can go to lunch.

Rob Wenger: She, she doesn't. She works hard on other things because she's that person. But yes, I agree.

Joanna Pineda: Are there other cool things that you can share?

Rob Wenger: It was all on this line. So there was a lot of stuff in sales. We use a lot of tools like gong is a tool we use. We record demos and gong had some really great AI tools that were introduced then, and they've been made a lot better since where we can get summaries and we put those summaries into Salesforce so that when we looking back at the account that we can see.

You know, what was discussed, what the key points, what their pain points are, make sure that we're meeting them with what we're proposing, things like that. There were a lot of things around that. It was really taking the features of existing software. So none of this during AI October was anything we developed.

These were all features of products we already had or things like, you know, Copilot that we added for it that then we just started building on. And then the sharing of the experience was To get others to see what they could be doing, right? [00:15:00] I mean the stuff I did was pretty basic. I didn't, well it wasn't part of the thing, but it was like, I just talked to ChatGPT all day long.

I don't ask it things that are specific about clients, but I ask it all kinds of questions about things I'm thinking about doing for the products, for example. And It's really good at telling me these things. So, no, it's been awesome. Like, you can go in and just ask it, what does an association do? And then drill down for hours and find out every role, every person in every association has.

And, you know, one of the things that we're trying to do with our product then is say, okay, there's You know, if you think about jobs to be done, there's a thousand jobs to be done in an association. Can we automate 200 of them? If we can, you know, that's 20 percent of their time they get back to do more strategic things or talk directly to members or whatever it is that they want to do to further their missions or run their organizations.

Joanna Pineda: So when you say you talk to ChatGPT all day long. As you're going about your day, you'll say, let me ask ChatGPT this, or do you set aside time to have these [00:16:00] conversations? And have you named your ChatGPT?

Rob Wenger: I name all of my AIs Samantha because of the movie Her. And if anybody's Ah, yes. That is a great look at what This is going to look like in a year or two, but anyway, yes, now I do.

I actually have this little cool device. You can't really see it, but it's this little thing called the stream deck. And it's got all these programmable buttons and I have a ChatGPT button. So when I push it, it pops up in front of me and no matter where it is on my massive number of screens and windows that are running, cause I can't, I just lose it, but yeah, I use it all day long.

Like you said, in the beginning, I. I have a question and I immediately ask it. By the way, I think Google search, it's dead to me. I think it's probably dead, a dead man walking kind of thing because I ask ChatGPT everything. The only thing I Google is when I need to go someplace, an address, right, not a physical address, but a web address, and I don't know what it is or I want to type it.

I type it into the, you know, the search bar and it comes up with it. But otherwise, ChatGPT is everything. And the reason I bring that up [00:17:00] is because it is sort of a cautionary thing. I do think. Traffic driven to websites is going to really, if app rate might be too strong, but it's going to be significantly impacted by people using ChatGPT rather than Google.

Because Google drives most traffic to most websites. Right, right. And if people aren't Googling things, ChatGPT doesn't currently, I mean they have The sort of rag, here's where I got the information from. But I find that I never click on that. I've been using it constantly for over a year and I almost never click on what they tell me to click on.

So, be aware of this, like, sort of lack of traffic and think about your marketing plans are going to have to be different.

Joanna Pineda: Right. Wow. Well, this is a very, very interesting and very intentional playbook for really kind of creating an AI adoption strategy at an organization. But let's turn to higher logic, right?

So you've got this suite of products. How are you integrating AI [00:18:00] into the products, and why are you integrating AI into the products?

Rob Wenger: Yeah, so we're being very intentional about it. I think what AI is great at, like, Unparalleled great at is understanding people's language, right, talking. It doesn't have to be English, whatever language of what you're actually asking in, you know, human spoken language is so good at understanding it.

Now, the jury's out on how good it is in giving you back what you're asking for, but in understanding it's really good. So that's one of the things. And so we're, we're really driving toward things that are going to move the needle on what do our organizations do. The first thing we actually built was a little chat window inside of our marketing product and forms, where a person can say, I want to send an email, you know, to my members.

Can you help me draft it? It's just a very, I could have done that in ChatGPT and pasted it in here, but we want to put that in place. And that's very easy because you can. Click it or not click it. It writes it for you and then you can edit it before you [00:19:00] send it. So it's, you know, an easy way to get into it.

Second thing we did is we did this thing called the bulk uploader, which I've wanted to do forever, but couldn't figure out how, which is we have a resource library, right? It has thousand PDFs in it. I wanted to upload all thousand PDFs at one go. It requires a, like a description and a title and potentially tags in order to arrange it, right?

With, ah, we could easily do that. Say, read this PDF and give me a summary on it. So again, pretty simple. Worst case, it's not exactly right. You click in, you say, oh, this is, you know, this is what it's about. So it's not the end of the world if it hallucinates. It doesn't, really, we found. Especially when, over time, we've upgraded the model.

I think our overall goal though is to automate, right? So having, there's two other things that I would mention. One, we just today launched in beta our, we call it a rag bot. It's an AI assistant, right? So it uses. The theory of retrieval augmented generation to take what the model has and augment it with what the association has as [00:20:00] private information.

And the private information is used to answer the question. The public and the private is combined to answer the question. So none of that private information goes to a model to learn. It immediately forgets it. But what we do is we combine search with, okay, here's what they're asking. Let's see if we have any documents that answer it.

We return those documents and then the AI just reads it. It summarizes it and then points to them and then forgets, right? So that's the biggest one, like I said, beta launch today. We have 15 beta clients on a call with us today, starting it up. It's available to any of our clients as of today, because you can just say, I want to be in the beta, but we'll probably put it out for, you know, like everyone should turn it on in the first week of March.

And we've been beta-ing it. Alpha and Beta with IBM for about four months now. IBM is our biggest client. They also make the AI tools, right? So we built it on top of Watson X, which is one of their big AI systems. And it's been really great. And so the feedback has been amazing. Four months in, in [00:21:00] Alpha.

Let's call it with them. So we're kicking off the general beta. So

Joanna Pineda: I want to make sure I understand what this thing does. You're saying, let's take a standard LLM, a large language model, and then using the RAG model, really train it in particular on, say, an organization's research reports. Train isn't the right word, but kind of prioritize the data, right, and have access to the private information from the association so that now the members can have a conversation with this chatbot that includes the data from the research and it also understands the association's acronyms, for example, because there's a jillion ASAs and a jillion NASs and, you know, and, you know, all that stuff.

Rob Wenger: So, and the reason I point out we don't train is because it doesn't remember, right? So it's, you know. That's right. But yeah, generally, you're right. And instead of saying data, I might say information. The difference being it's not private information. It's not people's email addresses and names and addresses and phone numbers and stuff like that.

It's really just the [00:22:00] content that's created. So the Q&A, the blogs, the uploaded resources, uploaded contract samples, whatever they put into their community is available to the search. Search runs, brings back the results, and then the AI reads the results of the search and summarizes the answer to the question.

It also, by the way, behind the scenes, does a whole lot of back and forth. So it'll appear to you, and ChatGPT works this way as well, it'll appear to you like you ask a question and it gives you an answer. What really happens is you ask a question, and a whole bunch of back and forth happens with different systems.

Right. And then eventually you get an answer that we've sanitized and said, okay, this looks like the right answer. So it's more complex than you think, more complex than I thought going in. But after four months of tuning it and prompt engineering and all kinds of feature development to make sure it's good and accurate, that's what we came out with.

So that's number one. Number two, our main strategy really is to automate as many of the sort of the, the [00:23:00] small repetitive tasks that have to happen to run an organization like an association. And anything that a computer and an AI can do, let's have them do it. So that the staff doesn't have to do it.

Simple example, I asked Chia at GPT, what are the big questions that someone in the association membership department gets from members? And it came up with a list of like 50 questions. And I said, how many can an AI answer? And it was probably about half of them. An AI could just read the website and give the answer.

And so if a, if a member, instead of calling the association and asking the question, ask the chatbot, and the chatbot can give them the answer, it's done. The other half, it could also answer with a little bit of work. Like, if I said, what are the member benefits? Right? It would be able to answer that just from looking at the website, so it could tell me who that is.

If I say, when does my membership renew, it has to look up who I am and see when my membership renews and answer that question. And the third [00:24:00] level is, can you register me for the webinar next week? right? So that's a sort of what I call doing. So one is, one is reading, one is doing. When, when you get to the doing level, that's when it gets really, really cool.

And these things we call agents.

Joanna Pineda: The agentic experience. The

Rob Wenger: Agentic experience, exactly. We're making dozens of agents with hundreds of skills each. And then the AI. And we get back to what I said about understanding English really well or language really well, is when you ask, it can just match up what you're asking with its skills and it's much more like, we call it deterministic.

It's much more likely to do the thing right than if you just let it do whatever it wants to do.

Joanna Pineda: Now let's take a quick break from the conversation for a word from our episode sponsor, HigherLogic. Is your association struggling to keep your members connected and engaged? HigherLogic has partnered with associations for nearly two decades to help them demonstrate member [00:25:00] value, increase engagement, and support their missions.

Their comprehensive HigherLogic Thrive platform was built to help associations create a complete member ecosystem and leverage data to personalize their member experiences. With one easy to use system, you can support all your offerings from your online community to your marketing campaigns, to your job board, to your mentoring and volunteering program and more.

Plus, their built in AI features, automated workflows, and robust analytics will save you tons of time so you can focus less on manual tasks and more on high impact strategic work. HigherLogic Thrive has everything associations need to gain real time insights to keep members connected, engaged, and coming back for more.

Well, I got another question for you. So within the HigherLogic community itself, there's lots and lots of data about what members are interested in, what they're saying, what they're thinking. Are you developing tools [00:26:00] so that the associations can get some insight into what's keeping them up at night, and how do you balance data concerns?

Amanda, maybe this is for you.

Rob Wenger: Yeah, I'll let her answer the concerns question. I'm not concerned but then I have her Backstop. So I don't need to be the reason I'm not concerned though is because it's the data. It's the organization's data, right? So we're not giving them anything other than what they already should have access to we're just doing it in a way That they can you know digest it, right?

So I don't think of that so much as AI as I do analytics

Joanna Pineda: Oh, that's true. Okay,

Rob Wenger: so we actually have a strategy we call the three A's it's AI analytics and automation They're not independent things, obviously, but that's how we think of everything we're working on in those categories, right? So this is what I would say in analytics, but you're right.

All the interactions, we expect there to be more interactions. With members and the association because of AI because they can start to think of this AI [00:27:00] Assistant as someone a companion who sits next to them that can answer questions for them Now what's interesting about associations though is I originally when I was thinking about this was like this is gonna be a big threat to associations because You've got no traffic being driven.

You've got these AI chatbots like, like ChatGPT and Gemini being trained on all the information in the world that has access to. Like, why do I need to go to an association to get any information? Well, there's two reasons. One is that there is all that private information, right? The stuff behind the file, right?

So you've got your journal, your spec, your standard, all the conversations, all the blogs, all the official publications of the organization. Exactly. So you've got all this content. that it can't get to, so it can't learn from. Even stuff like textbooks, in theory, are copyrighted, so it can't know that stuff, right?

So associations still have all the knowledge. And the second one is associations are the only place where you can continue to develop this knowledge, right? So I think of it as like, The, the best things associations have done [00:28:00] is gathered all of the world's experts in that thing in one place, right? So I started thinking about this with community, and this is the reason for community, right?

If I have a community for my association, my members can ask all the other members a question, right? So you're a junior member or you just have a hard question. You hope someone has that experience. You post in the community, the person has that experience or that, that expertise answers the question and it's great, you've got your answer.

The future now is, though, that now the, the bot could answer that question. Not ChatGPT, but the association's bot can answer that question. And so, there's a lot of value that's been created over all these years between all of the different content properties that an association has that you put into a bot, and now it's much easier to get that information out than it is to search.

But then, what happens if you can't get it, right? Because no one's asked the question before, but some spread is out there. So we had this cool feature, I think this is Awesome. It was my idea. So, you know, obviously I think it's awesome, but it's a button at the bottom that says, [00:29:00] not getting the answer you need, ask your fellow humans.

The idea of it is the bot goes and does its search and it tries to find an answer and it can't find one and what do you do? You push that button and it drafts a post to the community of basically what you were conversing with it with. It makes a summary and asks the community. This

Joanna Pineda: makes a phone call to a, to a human.

Rob Wenger: Phone's a friend. Phone's a friend. Phone's a friend is a good word. I might have to change it. Phone a friend.

Joanna Pineda: So, Amanda, let me ask you a question. So I do have some clients and I've heard from some association execs that are concerned about member data. Rob doesn't seem concerned. Are you concerned? And what safeguards are vendors like HigherLogic putting in place?

Amanda DeLuke: Yeah, so I think, again, it kind of goes back to the, the data classification and understanding what data is going to go into the model, and if it's even, you know, storing that data, if it's, right, it just, it's able to just, you know, go away and [00:30:00] you can ask a question and doesn't store it, so that's going to be really important, so keeping everything private, and the other, I would say, probably just the, the data classification and making sure it's a private model, it's not training your data.

Private model, right. Right.

Joanna Pineda: Right.

Amanda DeLuke: Yeah. And then also like some other things just to consider is just to make sure that when you are using AI, you're adhering to your current process. Like you shouldn't have to be changing your process. You should just be able to adhere to whatever applicable laws there are and, and whatever you, whatever you have.

For guardrails for your vendors, make sure that that's all in place. And if it is, then you can use it safely. I mean, this is one of my favorite features. I get so excited when you talk about this, Rob, because I'm like, this is a perfect use case because being able to leverage this data is really just going to allow in automating things is really going to allow.

Associations that have to wear so many hats be able to focus more on the human elements that are important to the business, [00:31:00] right? So I think all of this can be done very safely as long as again, because associations have different types of data that they're dealing with, it's just making sure that you identify the types of data.

Do your due diligence, making sure that it's private and it's not training a model and that it's staying internal. And then as long as you follow all of those guardrails, I think you can really do this safely. As well as like, if you do have something that you have concerns about, make sure to talk to your teams about that.

And then also you may just need to be transparent. Talk about in your terms of use or in your privacy policy, like, this is what's being used and just being very transparent and upfront. I think that's really important too.

Joanna Pineda: So what I'm hearing from you is have a conversation with your vendors or your industry partners to say what's going to happen to my data.

Is it private? You know, show me your terms of use and your, and your privacy policies so that I can be comfortable that we're not training the LLMs on my member data. Yeah.

Rob Wenger: What makes it easier [00:32:00] and harder for us, easier because there's a thing called SOC 2, right? Which is something we have to comply with.

Well,

Joanna Pineda: Explain what SOC 2 is.

Amanda DeLuke: Yeah, the way you explain it, it's, it's basically taking a look at all of the controls that you have in place at your business, right? And making sure they're, it's showing proof and accountability of those controls being in place and that they're working properly. So that's usually, you know, how I would describe SOC 2 audit.

And there's different types of SOC 2 and type 1, type 2, type 3, right? So depending on the type of data you're processing, but yeah. You want to be testing those controls to make sure they're working properly.

Rob Wenger: Yeah, so the good part about Suck2 is that because the guardrails have been set up over a period of many, many years, probably a decade now, I think, we have good checks and balances and we have people like Amanda who make sure we follow it.

It's a pain for me because I always want to violate them, not for real, like I don't want to purposely violate them, but I want to do something that [00:33:00] she says. No, you can't do it, which is great, but you know, it both speeds us up and slows us down. So an association probably doesn't need to and doesn't want to take on the burden of becoming SOC 2 compliant, but their vendors should, right?

So most of the organizations I think in the association space are certified SOC 2 compliant in one level or another. As long as they have that, they shouldn't be too worried about it, but they might want to be informed about it.

Joanna Pineda: Well, Rob and Amanda, I could talk to you guys all day, but before we go, what's two things that you would recommend to an association exec who says, I'm just getting started?

What should be in their playbook?

Rob Wenger: Yeah, I think number one is what we talked about earlier. You have a plan for how to deal with it. Even if you're not going to do it right away, make the plan. Because then you, you know what it's going to entail. The process of the planning will get you where you want to go.

And the thing I tell everybody, whether they're an association executive or my mom, is just use one of these tools. Play with [00:34:00] it. Like I said, I replaced Google with it, and I never looked back. It wasn't overnight, though. You know, a month ago, I would start to Google something, not get an answer, and be like, wait, why am I doing this?

It takes some time to build those muscle memory, so I would say, encourage your staff to just take that journey at least, and then you can become more sophisticated. I would also say, and I'll let Amanda talk more about this, but when we finished October, At the end of it, we put on a webinar for the association world to just like hear what we'd done and, and all of the sort of artifacts of that, the rules and regulations and stuff that Amanda and her team produced, we made available.

So if you do want to start, I'm not suggesting necessarily doing it in October, especially because I wouldn't wait till October, but if you can do, you know, April. Those resources are out there from us or there, there's a lot of other sources and Amanda, you're the expert on that. So,

Joanna Pineda: yeah, Amanda, you got the final word.

Amanda DeLuke:

So [00:35:00] clear company statement and communication is really important. Like, why are we doing this? Like, what is your scope? I think being very clear on how you communicate, what is important? Why do we want to use this AI and how are we going to benefit from it? So I think just answering those questions up front is really important, and it gives everyone sort of the fire to want to start using it and getting really excited.

So that's something that I felt when we started using AI October. I'm like, Oh my gosh, we have something clear coming from Rob, and we were all excited about using it. So I think that's really important. That's one step. And then really, I'm going to just hone in again on that data classification.

Understand the types of data that you're holding and the vendors processing it. Cause I've, I've seen some of these vendors, it's sort of happening like with this Trojan horse thing where AI just gets turned on by default and you don't know about it. I've seen that happen where you have to go in and actually manually turn it off.

Or I think that's really important is to be able to know your data. [00:36:00] Know where the data is being processed and know your vendors processing it, especially as it relates to AI.

Joanna Pineda: Wow. Well, I have enjoyed myself tremendously during this interview. I've learned a whole lot too. I want to thank you both, Rob and Amanda.

And thanks to ASAE for letting me host and being part of this amazing conversation.

Rob Wenger: Thanks for having us. It was a lot of fun being here and yeah, happy to answer any questions anytime.

Joanna Pineda: Thanks to everyone for listening to this episode of Associations NOW Presents. Join us each month as we explore key topics relevant to association professionals, discuss the challenges and opportunities in the field today, and highlight the significant impact that associations have on the economy, the U.

S., and the world. Again, we'd like to thank our episode sponsor Higher Logic. For more information, visit them online at higherlogic.com. Be sure to subscribe to our podcast on Apple, Spotify, or wherever you listen to your favorite podcast. And for more information on AI in the association space and how it's transforming our [00:37:00] community, visit associations now online at associationsnow.com.